Genomic differences among carriage and invasive nontypeable pneumococci circulating in South Africa

Abstract

Most pneumococci express a polysaccharide capsule, a key virulence factor and target for pneumococcal vaccines. However, pneumococci showing no serological evidence of capsule expression [nontypeable pneumococci (NTPn)] are more frequently isolated from carriage studies than in invasive disease. Limited data exist about the population structure of carriage NTPn from the African continent. We aimed to characterize carriage NTPn and compare them to previously described invasive NTPn. Carriage and invasive NTPn isolates were obtained from South African cross-sectional studies (2009 and 2012) and laboratory-based surveillance for invasive pneumococcal disease (2003–2013), respectively. Isolates were characterized by capsular locus sequence analysis, multilocus sequence typing, antimicrobial non-susceptibility patterns and phylogenetic analysis. NTPn represented 3.7 % (137/3721) of carriage isolates compared to 0.1 % (39/32 824) of invasive isolates (P<0.001), and 24 % (33/137) of individuals were co-colonized with encapsulated pneumococci. Non-susceptibility to cotrimoxazole [84 % (112/133) vs 44 % (17/39)], penicillin [77 % (102/133) vs 36 % (14/39)], erythromycin [53 % (70/133) vs 31 % (12/39)] and clindamycin [36 % (48/133) vs 18 % (7/39)] was higher (P=0.03) among carriage than invasive NTPn. Ninety-one per cent (124/137) of carriage NTPn had complete deletion of the capsular locus and 9 % (13/137) had capsule genes, compared to 44 % (17/39) and 56 % (22/39) of invasive NTPn, respectively. Carriage NTPn were slightly less diverse [Simpson’s diversity index (D)=0.92] compared to invasive NTPn [D=0.97]. Sixty-seven per cent (92/137) of carriage NTPn belonged to a lineage exclusive to NTPn strains compared to 23 % (9/39) of invasive NTPn. We identified 293 and 275 genes that were significantly associated with carriage and invasive NTPn, respectively. NTPn isolates detected in carriage differed from those causing invasive disease, which may explain their success in colonisation or in causing invasive disease.

Data Summary

The genome assemblies of all 137 carriage nontypeable pneumococcal isolates sequenced in this study have been deposited in GenBank under BioProject ID PRJNA289243. Individual sample accession numbers can be found in Table S1 (available with the online version of this article).

Impact Statement.

Most pneumococci are surrounded by a capsule, a key virulence determinant and a target for current pneumococcal vaccines. However, pneumococci showing no serological evidence of capsule expression (nontypeable) have been identified globally, mainly in carriage studies and rarely in invasive disease. There is a paucity of data about the population structure of carriage nontypeable pneumococci (NTPn) from the African continent. Using whole-genome sequencing, we characterized carriage NTPn from South Africa and compared the genomes to previously described invasive NTPn from South Africa. We show that, in contrast to our invasive NTPn, carriage NTPn had higher rates of antimicrobial resistance, the majority had genetically similar backgrounds belonging to a lineage exclusive to nontypeable strains and lacked conventional capsule genes. We also identified several genes that may play an important role in capsule-independent survival mechanisms of invasive NTPn. Continued surveillance and characterization of NTPn is important, since these isolates represent a reservoir of antibiotic-resistance genes for Streptococcus species and because the current pneumococcal vaccines are not effective against NTPn.

Introduction

Streptococcus pneumoniae is a respiratory pathogen that frequently colonizes the nasopharynx, and can cause significant morbidity and mortality [1]. Most pneumococci are surrounded by a capsular polysaccharide, a key virulence factor [2]. The capsule induces protective antibodies and forms the basis for current pneumococcal vaccines [3]. More than 100 capsular types can be identified by the Quellung reaction using serotype-specific anti-sera [4–7]. Some pneumococci show no serological evidence of capsule expression and these nontypeable pneumococci (NTPn) are predominantly carried in the nasopharynx [8–11]. The mechanisms underlying their nontypeability are diverse [8, 12–16]. NTPn have been associated with conjunctivitis outbreaks and non-invasive disease episodes, but rarely with invasive disease [17–19].

NTPn can be classified into two groups based on the gene content of the capsular polysaccharide synthesis (cps) locus [8]. In group I there are at least partial cps genes, whereas in group II the cps genes are completely deleted and may be replaced with non-cps genes. Group II is subdivided into four null capsule clades (NCCs) [19, 20]. NCC1 has the psk gene, NCC2 has aliC and aliD genes either with (NCC2a) or without (NCC2b) a putative toxin–antitoxin system (encoded by ntaAB genes). NCC3 has the aliD gene, and NCC4 contains only transposable elements in the cps locus.

Pneumococcal carriage studies from Europe reported specific lineages associated with NTPn strains [8, 11, 21]. A lineage of exclusively NTPn strains (classic lineage), comprising mainly sequence types (STs) 344, 448 and 449, and a lineage that appears to be related to encapsulated pneumococci (EcPn) strains (sporadic lineage) have been described [8, 21]. This was recently confirmed by whole-genome sequencing of 131 carriage NTPn isolates from 17 countries [22]. Carriage NTPn strains from Africa were not represented in these studies, and little is known about their population structure.

We previously characterized NTPn causing invasive pneumococcal disease (IPD) in South Africa [13]. These NTPn isolates represented 0.1 % (39/32 824) of IPD cases and 5.1 % (2/39) of IPD individuals were co-infected with EcPn. They were predominantly group I (56.4 %) isolates that harboured a variety of mutations within their cps locus and had higher antimicrobial non-susceptibilities than EcPn. Invasive NTPn isolates were diverse [Simpson’s diversity index (D) of 0.97; 95 % confidence interval 0.95–0.99] by multilocus sequence typing (MLST) analysis and the majority (79.4 %) belonged to sporadic lineages. In this study, we characterized carriage NTPn and compared their population structure to the previously described invasive NTPn from South Africa.

Methods

NTPn isolates

Invasive NTPn were obtained from active, national laboratory-based surveillance for IPD in South Africa from 2003 to 2013 and have been previously described [13]. Carriage NTPn were obtained from cross-sectional community carriage studies assessing the impact of seven-valent pneumococcal conjugate vaccine (PCV7) on pneumococcal nasopharyngeal colonization in South Africa. These cross-sectional carriage studies were conducted among randomly selected households with at least one child<2 years of age in a rural community (2009 and 2011) [23], and among human immunodeficiency virus (HIV)-infected and HIV-uninfected mother–child pairs in an urban community (2010 and 2012) [24]. PCV7 was introduced into the national immunization programme of South Africa in April 2009 and was replaced with PCV13 in July 2011.

Nasopharyngeal swabs (Medical Wire and Equipment) were collected, and placed in skimmed milk, tryptose, glycerol and glucose (STGG) broth transport media and stored at −70 °C. Swabs were cultured on 5 % horse blood agar with 5 mg gentamicin sulphate ml⁻¹ (Diagnostic Media Products, National Health Laboratory Service, Johannesburg, South Africa) and incubated at 37 °C in 5 % CO₂ for 48 h. Pneumococcal isolates were identified by susceptibility to ethylhydrocupreine hydrochloride (5 µg optochin; Becton Dickinson Microbiology Systems), bile solubility as well as real-time PCR detecting lytA [25]. Nontypeable pneumococcal isolates were detected during routine serotyping using the Quellung reaction (Statens Serum Institut, Copenhagen, Denmark) [26]. Co-colonization was defined as simultaneous isolation of NTPn with EcPn from the same nasopharyngeal swab.

Antimicrobial-susceptibility testing

Minimum inhibitory concentrations (MICs) for all antibiotics were determined by the broth microdilution method and interpreted using the Clinical and Laboratory Standards Institute 2014 MIC breakpoints [27]. Isolates defined as either intermediately resistant or resistant to any of the antibiotics were regarded as non-susceptible. For penicillin and ceftriaxone, the more conservative meningitis MIC breakpoints were used as we wanted more resolution and as this analysis was not to guide clinical therapy. Multidrug resistance was defined as non-susceptibility to three or more classes of antibiotics.

Nucleic acid extraction, genome sequencing and assembly

DNA extraction, whole-genome sequencing and genome analysis were carried out as previously described for invasive NTPn [13]. Briefly, genomic DNA was extracted from overnight broth cultures using a QIAamp DNA mini kit (Qiagen). Paired-end libraries (2×300 bp) were prepared using the Nextera XT DNA sample preparation kit (Illumina) and sequencing was performed on an Illumina MiSeq. Reads were imported into CLC Genomics Workbench v8 (CLC Bio) and were quality trimmed using the Trim sequences module in the software with the limit set at 0.05 (e.g. sequences with phred score of <30 were trimmed). The trimmed reads were assembled using the de novo assembly algorithm of CLC Genomics Workbench v8. The word size was set at 21, minimum contig length at 500 and bubble size at 80 % of mean read length. The assembled genomes were ordered relative to S. pneumoniae ATCC 700669 using Mauve [28] and annotated using Prokka v1.11 [29] (see Table S1 for genome characteristics).

Characterization of the cps locus and STs

The annotated genomes were imported into CLC Genomics Workbench v8. Using the ‘Find annotation’ tool, we searched for dexB and aliA genes, and extracted sequence data between these two genes (e.g. the cps locus) using the ‘Extract Sequences Tool’. STs were derived from genome data using the Bio-MLST-Check module (https://metacpan.org/release/AJPAGE/Bio-MLST-Check-1.133090). The eBURST v3 algorithm (http://eburst.mlst.net) was used to examine genetic relationships between carriage and invasive NTPn. A clonal complex (CC) was defined as isolates sharing 6/7 MLST alleles [30].

In silico prediction of ancestral serotypes for group I carriage NTPn

To predict ancestral serotypes of group I carriage NTPn (serotypes from which NTPn may have originated), three methods were used, namely serotype-ST associations, SeroBA and phylogenetic whole-genome analysis. In the first method, using the S. pneumoniae MLST database (http://pubmlst.org/spneumoniae/, accessed May 2016), we identified serotypes commonly associated with the ST of our group I carriage NTPn and, secondly, we used SeroBA v1.0.0, a rapid high-throughput in silico serotyping method for S. pneumoniae from whole-genome sequence data [31]. Lastly, we performed phylogenetic comparison (described below) of the carriage NTPn with invasive EcPn circulating in South Africa (Table S2). If group I NTPn clustered closely with a particular serotype in the phylogenetic tree, that serotype was assumed to be the likely ancestral serotype. If there were conflicting ancestral serotypes identified by the different methods, ancestral serotype assignment was based on the phylogenetic analysis result as this method accounted for the background genome of the isolate and was, therefore, assumed to represent the strain from which the NTPn isolate was potentially derived.

Molecular basis for the nontypeability of group I carriage NTPn

The ‘alignment tool’ in CLC Genomics Workbench v8 (default settings) was used to align the cps locus of each isolate, together with a number of its predicted ancestral serotype cps loci (Table S3). This was done to determine the genetic variation within the cps locus of group I carriage NTPn that may be responsible for the phenotypic nontypeability.

Phylogenetic analysis

To determine genetic relationships among the South African and other globally available carriage NTPn genomes, a rapid large-scale prokaryote pan-genome analysis pipeline (Roary) [32] and randomized accelerated maximum-likelihood (RAxML) analysis [33] were used to reconstruct a maximum-likelihood phylogenetic tree based on core genome SNPs of carriage NTPn from this study (n=137), and a previously described collection of carriage NTPn from other countries (n=131) [22]. The core genome alignment module in Roary [32] was used to extract predicted coding regions from Prokka-annotated assemblies and convert them to protein sequences. All protein sequences were compared with each other using blastp. Proteins that had alignment similarity of ≥70.0 % and were present in at least 90.0 % of the isolates were defined as the core genome. RAxML [33] was used to create a bootstrapped maximum-likelihood phylogenetic tree from the resulting core-genome alignment and visualized in Phandango v1.3.0 [34]. Invasive EcPn from South Africa (n=48) (Table S2) and previously described invasive NTPn from our setting [13] were included for comparison.

Genomic comparison of carriage and invasive NTPn

The pan-genome of carriage and invasive NTPn was inferred with Roary [32]. Prokka annotations were provided to Roary as an input and, in turn, Roary produced a gene presence/absence matrix for the isolates. To identify genes that were associated with carriage or invasive NTPn, genes were classified as unique if they were present in at least one carriage NTPn isolate and absent in invasive NTPn isolates or vice versa. We also examined whether genes were significantly over-represented in carriage versus invasive isolates and vice versa, using the chi-square test with significance assessed at P<0.05. wego software [35] (http://wego.genomics.org.cn/, accessed 1 February 2017) was used to perform gene ontology functional classification for the genes that were significantly over-represented in invasive NTPn and present in the majority (>70 %) of invasive NTPn isolates.

Statistical analyses

Proportional differences were analysed using the chi-square or Fisher’s exact test, where appropriate. Statistical analyses were performed with GraphPad InStat 3 (GraphPad Software) and Microsoft Excel. ST diversity was calculated using Simpson’s diversity index (D) (http://darwin.phyloviz.net/dokuwiki/doku.php – under the comparing partitions tab). D ranges from 0 to 1, and values closer to 1 indicate higher diversity. Statistical significance was assessed at P<0.05.

Results

Prevalence of carriage NTPn

The overall prevalence of S. pneumoniae colonization was 31.7 % (3721/11 739), for which 3.7 % (137/3721) isolates were NTPn. Amongst participants colonized with NTPn and for whom demographic data were available, 60.6 % (63/104) were female, 81.5 % (110/135) were children aged<5 years and 43.6 % (41/94) were HIV infected. A total of 24.1% (33/137) of individuals carrying NTPn were co-colonized with EcPn, and the most common co-colonizing serotypes were serotype 19F (5/33, 15.2%), 6B (4/33, 12.1%), 6A (3/33, 9.1%), 23F (3/33, 9.1%) and 34 (3/33, 9.1 %).

Antimicrobial non-susceptibility of carriage NTPn compared to invasive NTPn

Among carriage NTPn, 84.2 % (112/133) were non-susceptible to cotrimoxazole, 76.7 % (102/133) to penicillin, 52.6 % (70/133) to erythromycin, 51.1 % (68/133) to tetracycline and 36.1 % (48/133) to clindamycin (Fig. 1). Antimicrobial-susceptibility testing was not performed for four isolates due to loss of viability during storage. Overall, with the exception of chloramphenicol non-susceptibility, which was higher in invasive NTPn compared to carriage NTPn [10.3 % (4/39) versus 0.8 % (1/133), P=0.01], non-susceptibility was higher among carriage NTPn than invasive NTPn for cotrimoxazole [84.2 % (112/133) versus 43.6 % (17/39), P<0.01], penicillin [76.7 % (102/133) versus 35.9 % (14/39), P<0.01], erythromycin [52.6 % (70/133) versus 30.8 % (12/39), P=0.02] and clindamycin [36.1 % (48/133) versus 17.9 % (7/39), P=0.03]. Multidrug non-susceptibility was higher in carriage NTPn compared to invasive NTPn [58.6 % (78/133) versus 30.8 % (12/39), P<0.01].

Fig. 1.

Antimicrobial non-susceptibility of carriage (2009 to 2012) and invasive nontypeable S. pneumoniae (2003 to 2013) from South Africa.

Carriage NTPn cps loci

Nine per cent (13/137) of carriage NTPn had conventional cps loci and were classified as group I (Table 1). For 84.6 % (11/13) of group I carriage NTPn, all methods used to predict ancestral serotypes were in agreement. Two isolates (NT66 and NT103) had conflicting results where the ancestral serotype predicted by the capsular genes differed from that of the ST-associated serotype and phylogenetic clustering, possibly due to serotype switching. The most common predicted ancestral serotypes were serotype 3 (4/13, 30.8%) and 13 (2/13, 15.4 %). Compared to their predicted ancestral serotypes (or switched serotype locus for NT66 and NT103), group I carriage NTPn had SNPs and deletions within their cps loci (Table 1). The remaining carriage NTPn (124/137, 90.5 %) were classified as group II and had pspK (6/124, 4.8 %), aliC, aliD and ntaAB (104/124, 83.9 %), aliC and aliD (5/124, 4.0 %), aliD (3/124, 2.4 %), transposable elements (3/124, 2.4%) and aliC (3/124, 2.4 %) in place of the cps locus. Based on the classification by Park et al. [20, 36], group II carriage NTPn were defined as NCC1, NCC2a, NCC2b, NCC3 and NCC4. Three group II carriage NTPn (3/124, 2.4 %) contained only aliC and did not belong to any previously defined clade.

Table 1.

STs, capsular loci (cps) and ancestral serotypes of group I carriage non-typeable S. pneumoniae (n=13) in South Africa, 2009–2012

Isolate no.	ST	ST associated serotype	SeroBA serotype	Phylogenetic clustering	Predicted ancestral serotype	Mutations in the cps locus relative to ancestral serotypes	Comments
NT185	700	3	3	3	3	Deletion of aliB, tnp, wzg, wzh. Partial deletion of wzd and wze. SNPs in wze, ugd, wchE and galU. Insertion in wze.
NT69	458	3	3	3	3	Partial deletion of aliB and pgm. SNPs in wzh and aliB.
NT89	458	3	3	3	3	Partial deletion of aliB, pgm and tnp.
NT205	458	3	3	3	3	Partial deletion of pgm.
NT191	4872	19A	19A	19A	19A	Deletion of IS1670. Partial deletion of IS1202. SNPs in IS1202, wzg, wze, wzx, rmlD. Insertion in IS1202, wzx and rmlD.
NT103	5410	4	19A	4	4	Deletion of IS1202. Partial deletion of IS1670 and rmlD. SNPs in wzg and IS1670. Insertion in rmlD.	Co-carried with 19A. Capsule compared to 19A*.
NT244	6277	19F	19F	19F	19F	Deletion of aliB and introns. Partial deletion and SNPs in tnp. Insertion in wzx.	Co-carried with 19F.
NT134	217	1	1	1	1	Deletion of aliB. Partial deletion of tnp and wzx. SNPs in tnp, wzd, wze and gla.
NT207	2285	6A	6A	6A	6A	SNPs in wzd. Partial deletion of wzx.
NT171	5647	13	13	13	13	Deletion of tnp. Partial deletion of whaG and wzy. SNPs in wciF.
NT177	393	38	38	38	38	SNP in wcil. Partial deletion of wzy, wcyD and gla. Insertion in wzy.
NT206	447	37	37	37	37	SNPs in wzg, wzh, wzd, wze, wchA, wciB, wzy, wciG, glf and wcjE
NT66	10 938	None	10A	13	13	SNP in aliB. Partial deletion of tnp.	Capsule compared to 10A*.

*Isolates had capsular genes not matching their predicted ancestral serotypes, maybe due to serotype switching (this is evident for isolate NT103, which was co-carried with serotype 19A and, therefore, likely took up the 19A capsule); hence, the capsule locus was compared to the switched capsule and not the ancestral serotype locus.

ST analysis of carriage NTPn and comparison to invasive NTPn

Among the 137 carriage NTPn, 44 STs were identified of which 24 were novel STs. The Simpson’s diversity index among the carriage NTPn was 0.92 (95 % confidence interval: 0.90–0.95) (Fig. 2). Among the 13 group I carriage NTPn, only 3 isolates belonged to the same ST (ST458) (Fig. 2, Table 1). Among 124 group II carriage NTPn, 33 STs were identified of which 23 were novel STs. STs 9810 (25/124, 20.2%), 344 (21/124, 16.9%), 9809 (13/124, 10.5%), 9811 (13/124, 10.5%), 10 972 (7/124, 5.6%), 9519 (6/124, 4.8%) and 7604 (5/124, 4.0%) represented 72.6 % (90/124) of group II carriage NTPn. Four CCs namely CC344 (49/124, 39.5%), CC9810 (33/124, 26.6%), CC10972 (21/124, 16.9%) and CC9519 (8/124, 6.5 %) accounted for 89.5 % (111/124) of group II carriage NTPn. The majority (82/124, 66.1 %) of group II carriage NTPn were related (sharing at least 5/7 MLST alleles) to international PMEN clones Norway^NT-ST344 and USA^NT-ST448 (http://www.pneumogen.net/pmen/). Six STs (ST217, ST344, ST9809, ST9810, ST9811 and ST10241) were identified in both carriage (54.0 %, 74/137) and invasive (41.0 %, 16/39) NTPn.

Fig. 2.

eBURST comparison of nontypeable S. pneumoniae (n=176) showing relationships between STs found in invasive (n=39) [13] and carriage (n=137) isolates from South Africa. STs shown in black were only found in invasive isolates, those shown in green were found in carriage isolates and those shown in pink were found in both. The size of each circle corresponds with the number of isolates. Clusters of linked isolates correspond to CCs (isolates sharing at least five of the seven alleles) and blue indicates the founding genotype. Group I carriage isolates are underlined with a dashed line. Asterisks denote new STs.

Phylogenetic analysis of carriage and invasive NTPn

Carriage and invasive NTPn isolates clustered by ST, nontypeable lineage and nontypeable group (Fig. 3). Within the classic lineage, two CCs, namely CC344 and CC9810, were dominant. Within CC344, ST344 strains from South Africa clustered more closely with single-locus variant ST9809 South African isolates than with ST344 isolates from other countries. Overall, 67.2% (92/137) of carriage and 23.1 % (9/39) (P<0.01) of invasive NTPn from our study belonged to the classic nontypeable lineage. The remaining carriage (45/137, 32.8%) and invasive (30/39, 76.9%) NTPn clustered with EcPn isolates in the sporadic nontypeable lineage. Group I carriage NTPn were closely related to their predicted ancestral serotypes (Table 1).

Fig. 3.

Maximum-likelihood phylogenetic tree of carriage (n=137) and invasive (n=39) nontypeable S. pneumoniae from South Africa, and a collection of carriage nontypeable isolates (n=131) from 17 countries (2 countries in Africa, 6 in Europe, 3 in North America, 1 in South America, 1 in Australasia and 4 in Asia). Encapsulated, invasive S. pneumoniae isolates representative of different serotypes from South Africa (n=47) were included for comparison. In CCs, other includes CCs: 74, 105, 353, 941, 1473, 4084, 5604 and 9817. The tree is based on core genome SNPs of the isolates. Bar, 0.04 nucleotide changes per 100 bp.

Genomic comparison of carriage and invasive NTPn

We identified a total of 11 959 genes in the pan-genome of carriage and invasive NTPn, of which 39.9 % (4777/11 959) encoded hypothetical proteins. The core genome for all NTPn isolates was 1102 coding sequences, which translated to 55.1 % of the total genome. We found 5059 and 1298 genes that were exclusively present in carriage and invasive NTPn, respectively (Tables S4 and S5). Of these, 43.3 % (2193/5059) and 41.7 % (541/1298) encoded hypothetical proteins, in carriage and invasive NTPn, respectively. We also found 293 and 275 genes that were significantly associated with carriage and invasive NTPn, respectively (Tables S6 and S7). Of the 275 genes significantly associated with invasive NTPn, 59 genes (highlighted in red in Table s7) were present in at least 70 % of invasive isolates, and 52 genes (88.1%) had blast matches to known proteins and were assigned to three gene ontology classes with 260 functional terms (Fig. 4). Assignments to biological processes, molecular function and cellular components were 48.8 % (127/260), 34.2 % (89/260) and 16.9 % (44/260), respectively. Within biological processes, metabolic processes were highly represented (26.9%, 14/52); and within molecular function, catalytic activity (38.5 %, 20/52) and binding (26.9 %, 14/52) were highly represented.

Fig. 4.

Gene ontology classification of 52 genes significantly associated with invasive nontypeable S. pneumoniae compared to carriage nontypeable strains. The genes were classified into three classes (biological processes, molecular function and cellular components) using wego software.

Discussion

We characterized and compared carriage and invasive NTPn circulating in South Africa. While NTPn caused IPD at a low prevalence (0.1 %) [13], they were more frequently isolated in carriage (3.7 %). Carriage NTPn were predominantly group II (90.5 %) and classic NTPn lineage (67.1 %), whereas invasive NTPn were predominantly group I (56.4 %) and sporadic NTPn lineage (76.9 %) . Non-susceptibility to cotrimoxazole, penicillin, erythromycin and clindamycin was significantly higher in carriage NTPn compared to invasive NTPn. Carriage NTPn isolates were slightly less diverse than invasive NTPn (D=0.92 vs 0.97). We identified 293 and 275 genes that were exclusively found in carriage and invasive NTPn, respectively.

A small proportion (9.5 %) of our carriage NTPn had partial cps genes compared to our invasive NTPn (56.4 %) (P<0.01) [13]. This is similar to a pneumococcal carriage study conducted in Thailand from 2007 to 2010, where 9.2 % (46/501) of carriage NTPn had partial cps genes [37]. For the two carriage group I isolates that had capsular genes not matching their predicted ancestral serotypes, we hypothesize that there may have been a serotype switch and that during or after this event the isolates lost the ability to express a capsule. In particular, isolate NT103 was co-carried with a serotype 19A isolate; therefore, we speculate that a serotype switch event resulted in the acquisition of 19A capsular genes.

As expected, we observed a higher prevalence (90.5 %) of group II carriage NTPn than invasive NTPn (43.6 %) (P<0.01) [13]. This was similar to the Thailand study, where group II represented 90.8 % (455/501) of carriage NTPn [37]. Three of our group II carriage isolates (2.4 %) contained only aliC in the cps region and, thus, did not belong to any clade using the current classification [19, 20]. We propose that cps with aliC be classified NCC5.

Twenty-four per cent of our carriage NTPn were co-carried with EcPn compared to invasive NTPn where 5.1 % were co-infected with EcPn [13]. The EcPn co-colonizing with carriage NTPn were diverse, but serotypes 19F, 6B, 6A, 23F and 34 were most common. One group I carriage NTPn isolate (NT244), with a predicted 19F ancestral serotype, co-colonized with a serotype 19F isolate. It is possible that NT244 is a variant of this 19F isolate; however, we did not assess the relationship between co-colonizing pairs. Schaffner et al. described a serotype 18C carriage isolate that lost its capsule, resulting in the co-existence of encapsulated and unencapsulated variants of the same strain [38]. We have also previously described two cases of IPD caused by co-infecting encapsulated and unencapsulated variants of serotype 1 and 18C [14].

As co-colonization of NTPn with EcPn occurs frequently and provides an opportunity for exchange of genetic material [9, 11, 39], NTPn can serve as reservoir of resistance genes. Evidence for this was a study in Switzerland that found a serotype 19F clone became resistant to penicillin by acquisition of a gene from NTPn [40]. Recently, using whole-genome sequencing, a study in Thailand showed that higher rates of DNA exchange occurred in NTPn compared to EcPn, and the most commonly exchanged genes were those associated with antimicrobial resistance and immune interaction [37].

There were significantly higher rates of antimicrobial non-susceptibility among carriage NTPn than invasive NTPn for cotrimoxazole, penicillin, erythromycin and clindamycin. Since the majority of carriage NTPn (90.5 %) lacked a capsule (true nontypeable), they are more likely to acquire resistance genes, compared to invasive NTPn where 56.4 % were encapsulated but at some point sporadically lost the ability to express the capsule. Also, carriage NTPn have a greater risk of antimicrobial exposure than invasive NTPn and are, therefore, more likely to acquire resistance genes [41]. Lastly, this high resistance in carriage NTPn may be driven by global multidrug-resistant nontypeable clones Norway^NT-ST344 and USA^NT-ST448, as two-thirds of our carriage NTPn were related to these clones. High rates of antimicrobial non-susceptibility are common among carriage NTPn isolates [11, 37, 39].

Compared to our invasive NTPn, which were slightly more heterogeneous [13], MLST analysis of carriage NTPn showed that this population is slightly less diverse. This may be due to the fact that the majority (two-thirds) of carriage NTPn isolates were of the same clones (Norway^NT-ST344 and USA^NT-ST448). Similarly, the population structure of NTPn isolates from carriage studies conducted in Portugal in children aged<5 years between 2001 and 2007 was also clonal, with 83 % (20/24) of NTPn isolates belonging to global clones Norway^NT-ST344 and USA^NT-ST448 [11]. In Thailand, approximately 40 % of carriage NTPn were represented by three STs: ST448, ST4133 and ST4136.

Immune deficiencies of the host and/or other known or novel virulence factors have been proposed as explanations for the ability of NTPn to cause IPD without the capsule [12]. However, in our previous study, the majority of patients with invasive NTPn infection (87.0%) were not immunocompromised, although the numbers were small [13]. The genes that we found to be significantly associated with invasive NTPn may play an important role in aiding NTPn to cause IPD and should be further investigated.

Our study has some limitations. It is possible that our NTPn isolates may have lost the ability to express a capsule during sub-culturing in the laboratory, but we believe that if this were the case we would detect NTPn much more frequently. We did not perform in vitro experiments to confirm whether the mutations in the cps locus of our isolates were responsible for the lack of capsule expression. However, in the majority of these isolates, there were complete deletions of cps genes making expression of a capsule impossible. We did not perform gene knockout or mouse model studies for the genes that were exclusively present in invasive NTPn compared to carriage NTPn to determine their role in virulence.

In conclusion, our study showed differences between carriage and invasive NTPn related to prevalence, capsule locus, relatedness, virulence, antimicrobial resistance and genome content. The genes that were significantly associated with invasive NTPn in this study should be further investigated to determine their role in virulence. Continued surveillance of NTPn is important since these isolates represent a reservoir of antibiotic-resistance genes for EcPn.

Data bibliography

1. Bentley SD, Aanensen DM, Mavroidi A, Saunders D, Rabbinowitsch E et al. Genetic analysis of the capsular biosynthetic locus from all 90 pneumococcal serotypes. PLoS Genet 2:e31 (2006).

2. Hilty M, Wüthrich D, Salter SJ, Engel H, Campbell S et al. Global phylogenomic analysis of nonencapsulated Streptococcus pneumoniae reveals a deep-branching classic lineage that is distinct from multiple sporadic lineages. Genome Biol Evol 6:3281–3294 (2014).

3. Mohale T, Wolter N, Allam M, Ndlangisa K, Crowther-Gibson P et al. Genomic analysis of nontypeable pneumococci causing invasive pneumococcal disease in South Africa, 2003–2013. BMC Genomics 17:470 (2016).